Automatic Distribution of PRVs for Leakage Reduction ^†

Abstract

In this work, an automatic distribution of the pressure reduction valves (PRVs) is proposed. First, a well-calibrated hydraulic model is required. The model of Manresa, a city of Catalunya in the Mediterranean area, was calibrated using pressure sensors, and the background leakage was estimated using weighted emitter coefficients. Simulating the model in real boundary conditions highlights the areas of maximum background leakage. The manual introduction of a PRV shows its effectiveness regarding leakage reduction. An algorithm for finding the high-pressure areas and their boundary pipes is presented. The introduction of the PRV, taking into account the flow constraints, produces a new scenario. Finally, the leakage reduction thanks to the pressure control by means of new actuators is evaluated. The leakage is reduced by around 6%.

1. Introduction

Leakage in a Water Distribution Network (WDN) is always a main issue. The community devotes continuous efforts to its reduction. Furthermore, during drought episodes, leakage reduction becomes imperative in some regions, like the Mediterranean region. In such situations, pressure management that leads to leakage reduction appears to be suitable both for its effectiveness and the general agreement in reducing the pressure service.

There is plenty of research regarding how to manage the pressure within a WDN to reduce leakage. In [1], the authors present a case study for improving the performance of a multi-PRV supply system in 2001. Lately, more and more sophisticated control strategies from predictive control [2] to distributed control [3] in case studies from all over the world [4] have been developed. Nevertheless, when a practitioner decides to reduce the pressure in a WDN, the introduction of new actuators (PRVs) must be well evaluated given the investment involved. Moreover, the study is highly site-dependent as the geography of the region and the topology of the network condition the decisions. Thus, the use of the hydraulic model and simulations help regarding decision making.

In this work, an automatic distribution of a PRV is proposed. First, a well-calibrated hydraulic model is required. In [4], the model of Manresa, a city of Catalunya in the Mediterranean area, was calibrated using pressure sensors, and the background leakage was estimated using weighted emitter coefficients. Simulating the model in real boundary conditions highlights the areas of maximum background leakage. The manual introduction of a PRV shows its effectiveness in terms of leakage reduction. An algorithm for finding the high-pressure areas and their boundary pipes is presented. The introduction of the PRV taking into account the flow constraints produces a new scenario. Finally, the leakage reduction thanks to the pressure control by means of new actuators is evaluated.

2. Materials and Methods

The starting point of the work is a real network. This network is presented as the case study in the first subsection. Within this network, there are different pressure levels that are not optimized in terms of leakage. First, the effect on leakage of a manual distribution of PRV is evaluated. This evaluation suggests an automatic procedure for PRV distribution that can be applied to any network when a hydraulic model is available. This method is presented in the third subsection.

2.1. Case Study

The WDN object of this work includes 39,000 consumers; its mean flow is 373 m³/h. Its model (Figure 1) includes 5123 nodes and 5285 pipes. It is fed by gravity from a tank where an electromagnetic flowmeter provides the global consumption online.

The model includes background leakage model given a performance of 85% [5]. Within the WDN, there are nodes with pressure from 25 mWC to above 60 mWC.

2.2. Manual PRV Distribution

The initial simulation of the model produces the pressure values for all the nodes. The valves will be introduced following two criteria. First, the pressure must be over the 40 mWC that is considered enough for satisfying the service. Furthermore, they are introduced where the flow cannot be easily diverted, increasing the pressure in other regions.

The manual procedure is applied iteratively, discarding those valves whose improvement in terms of pressure reduction is negligible and repeating it until there are no more clear possible locations.

This procedure improves the performance, as indicated in the Results section, but it is time-consuming and requires the contribution of an expert.

2.3. Automatic PRV Distribution

In order to utilize the information generated by the simulation model, a traversal algorithm is used for finding the areas of high pressure and their boundaries. The Algorithm 1 assigns to those nodes with a pressure over a threshold a label of the high-pressure zone where they belong. The boundary nodes between high- and normal-pressure zones are chosen for valve installation.

Algorithm 1: High-pressure zones definition

Input: Node ID; Nodes’ Pressure; Links’ nodes; Pthreshold 1 Choose a non-visited node i     If Pi > Pthreshold         Assign high-pressure zone Visited         2 Obtain consecutive nodes j               If Pj > Pthreshold                   Assign high-pressure zone Visited                   Go to 2                else                   Assign high-pressure zone and boundary                   Go to 1  Return: High-pressure zones and boundaries

3. Results

In the first manual iteration, 10 PRVs are introduced. This reduces the inflow by 0.59%; after two iterations, this reduction increases to 0.8%. The automatic algorithm improves this result without the expert information. The inflow reduction is 0.9%. The number of valves in the automatic procedure reaches 33. The manual study of so many valves appears to be unrealistic.

Table 1. Results of three manual iterations and the automatic approach in terms of leakage and pressure reduction.

Property	Without PRV	Manual 1st Iteration	Manual 2nd Iteration	Manual 3rd Iteration	Automatic
Number of valves	0	10	4	8	33
Inflow [L/s]	58.04	57.7	57.7	57.57	57.514
Inflow decrease [L/s]		0.34	0.34	0.47	0.522
Inflow decrease [%]		0.59	0.58	0.8	0.9
Mean Pressure [mWC]	49.78	48.72	48.74	48.05	47.21
Mean Pressure decrease [mWC]		1.06	1.04	1.73	2.57
Mean Pressure decrease [%]		2.13	2.09	3.5	5.16

presents the results of three manual iterations and the automatic approach in terms of leakage and pressure reduction. In Figure 1, the location of the third manual iteration and the automatic location are presented.

4. Conclusions

This paper presents an algorithm that automatically distributes PRVs, enabling reducing the pressure in a DMA. This pressure reduction produces a decrease in inflow due to the dependence of background leakage on pressure.

The results obtained by the automatic algorithm include the valves suggested by a manual iterative process led by an expert. The algorithm requires many PRVs that can be an excessive investment; thus, a valve selection procedure could be developed as future work. Nevertheless, this work demonstrates that the time-consuming process of finding all the relevant locations for the valves can be automatized.

The inflow decrease obtained is due to the leakage as the demands have been assumed to be all volumetric. The decrease of 0.9% in the inflow would be higher in a less efficient system (performance of 85%). In this scenario, this corresponds to a reduction of 6% regarding the leakage. Furthermore, there is a component of demand that is pressure-dependent [4], and this increases the actual water saving due to pressure reduction. Such demand reduction should be studied.